Вниманию пользователя 
akor168
Along America's west coast, the world's most valuable companies are racing to make artificial intelligence smarter. Google and Facebook have boasted of experiments using billions of photos and thousands of high-powered processors. But late last year, a project in eastern Tennessee quietly exceeded the scale of any corporate AI lab. It was run by the US government. "Deep learning has never been scaled to such levels of performance before," says Prabhat, who leads a research group at the National Energy Research Scientific Computing Center at Lawrence Berkeley National Lab. His group collaborated with researchers at Summit's home base, Oak Ridge National Lab. Fittingly, the world's most powerful computer's AI workout was focused on one of the world's largest problems: climate change.
И это еще неплохое применение, раньше National Energy Research Scientific Computing Center смоделировал бы какую-нибудь очередную водородную бомбу.
И это еще неплохое применение, раньше National Energy Research Scientific Computing Center смоделировал бы какую-нибудь очередную водородную бомбу.
I cannot quite tell whether you are joking or not.
In these applications they often use neural nets in conjunction with a model. It is true that it is typically less tractable analytically though.
Difference between what and what? Linear regression is not necessarily easy to understand either although it is more tractable.
Коэффициенты регрессии допускают самую непосредственную интерпретацию в терминах предметной области -- как вклады соответствующих переменных при неизменных остальных. Первейший sanity check факторной модели - это имеют ли они смысл. В противоположность этому, веса нейронной сети специалисту в предметной области не скажут ничего.
Well, this is true for some application but in many areas linear models are used with features which have little intrinsic meaning. An what about non-parametric methods? What about random forests or boosting? Neural networks are hard to interpret but so are many other classical methods. The difference is that neural netoworks are hard to analyze mathematically.